Variable Integral Adaptive Backstepping Hysteresis Admittance Compliant Control of Sensorless Single Joint Exoskeletons With Unknown Parameters

This paper proposes a variable integral adaptive backstepping controller (VIABC) and a hysteresis admittance model for compliant control of single joint exoskeletons widely used in robotic rehabilitation. The proposed method is realized via iterative adaptive backstepping design with the integral of the error added to the first stabilizing function. An integral law is then formulated to improve transient and steady-state tracking of the admittance reference trajectory in the presence of unknown parameters and uncertainties. The proposed VIABC obtains desired derivatives of input signals from the admittance model thereby avoiding the need for a derivative filter which could lead to noise amplification and increased computational cost. The proposed controller is tuned via genetic algorithm optimization on an identified model of the actuator using a fitness function designed to minimize the tracking error while preventing control signal chattering. The method is experimentally implemented using an inexpensive microcontroller in real time on a robotic exoskeleton attached to the shank of a clinical mannequin with unknown parameters while the interaction torque and velocity feedback signals are obtained from an extended state observer. The experimental results show that the VIABC performs significantly better than optimized conventional adaptive backstepping and PID controllers in tracking the admittance trajectory during the online identification phase and post-identification with gravitation and friction compensation. The results also show that the proposed method can be implemented for stand-alone operation using inexpensive hardware with low sample rates.